How to Make a SaaS Product Video with AI — Step-by-Step (+ 4 Prompts)

Steal our 4 prompts

June 24, 2026

🎥 Prefer watching over reading?

‍We’ve put together a full video breaking down all these concepts with real visual examples and case studies. Click the video below to watch the full episode!

AI has touched every part of video production. It helps with scripting, footage generation, and post-production. But most SaaS companies are using it wrong — not because the tools are bad, but because nobody explained where AI actually helps and where it quietly breaks things.

In this guide, I'm walking through the complete workflow we use at Lava Media to produce SaaS marketing videos with AI.

Before You Touch Any Tool: Define Your Format

The biggest mistake in SaaS video production isn't a production mistake — it's a strategy mistake. Before you open Claude, Runway, or any other tool, you need to know three things: who you're making the video for, what you're promoting, and what action you want viewers to take.

These answers completely determine your video format. Here's how it breaks down:

Scenario 1 — Audience: Investors. Goal: Pitching.

Keep it under 2–3 minutes. Focus on market opportunity — stock footage or AI-generated visuals work well here. Show your actual team on camera (this matters for credibility). Visualize your technology with 2D or 3D animation.

Scenario 2 — Audience: Potential users. Goal: Traffic and leads.

A series of short SaaS video ads works better than one long video. Each video covers one specific product benefit. Runtime: 6 to 15 seconds. Use AI-generated visuals for familiar, relatable scenes combined with 2D animation for the product interface.

Scenario 3 — Audience: Website visitors. Goal: Conversion.

Your SaaS product demo video should strengthen your highest-traffic pages and communicate your key value for people who don't read. Runtime: up to 60 seconds. AI visuals for lifestyle context, 2D or 3D animation for the product itself.

Get the format wrong and no amount of production quality will save the video. Get it right and even a simple production can convert.

Step 1: Competitor Research and Reference Collection

Once you know your format and goal, it's time to collect references. This step is how the visual direction of your entire video gets established.

The goal is to understand what already exists in your space and what you'll need to stand out from. Yes, you have brand guidelines. But an initial reference search shows you the landscape before you brief a creative team.

Here's how we do it at Lava Media using Claude or Gemini.

Prompt 1 — Competitor Research

I'm a [role] at [Company Name]. Website: [URL] Product: [brief description] Target audience: [who they are] Market/niche: [industry or category] Geography: [e.g. US, Europe, global] Task: Find 15 direct competitors. For each one provide: — Company name — Website URL — What they sell and to whom — Any notable differentiator Sort by relevance to my business.

Run this, quickly verify the results are actually your competitors (AI sometimes misreads the product category), and refine if needed. They don't have to be direct competitors — adjacent companies with strong video content are just as useful for references.

Prompt 2 — YouTube Channel Finder

Here is a list of companies: [paste list] Task: Find the official YouTube channel for each company. Return a table with two columns: — Company name — YouTube channel URL If no channel is found, mark it as "Not found".

Go through the list, open their channels, and build a shortlist of what you like — noting specifically what appeals to you and why. When we were preparing the brief for our client AMPD, we needed to visualize how their system navigates a decision path. The reference phase let us show the client a screenshot with a note: "we want this — an abstract 3D space with a movement line from one decision to the next." That one reference aligned the entire creative direction before any production started.

The style decision at this stage determines everything downstream. Studying the best SaaS demo videos in your space gives you a grounded answer — and shows you where there's room to differentiate.

Looking at what your competitors have already made is one of the fastest ways to figure out what you don't want to make.

Step 2: Writing Your Script with AI

With references collected and a clear creative direction, it's time to write the script.

Prompt 3 — Script Draft

You are a scriptwriter specializing in B2B video production. Write a video script based on the following: Product: [name and one-line description] Key benefits: [list 2–4 main benefits] Value delivered: [what problem it solves and for whom] Target audience: [job title, company size, industry] Core message: [the one thing viewers must take away] CTA: [what you want viewers to do after watching] Video type: [explainer / product demo / testimonial / ad] Runtime: [e.g. 30 sec / 60 sec / 2 min] Tone: [e.g. professional, conversational, bold] Format the output as: — Scene-by-scene breakdown — Voiceover / on-screen text — Visual direction per scene

Fill in every field specifically. The more precise your inputs, the better the output. A vague brief gives you a generic script that sounds like every other SaaS marketing video.

The Voice Memo Method

One approach that often works better: record yourself talking through the product. Cover the benefits, the user journey, the pain points — in your own words. Then use this prompt first to transcribe it before running the script prompt above.

Prompt 4 — Raw Transcription

Step 1 — Transcribe only. Below is my voice recording / raw notes: [paste text or attach audio] Transcribe it word for word. Do not clean up, summarize, rephrase, or remove anything. Preserve every word exactly as spoken, including filler words and repetitions. Output only the raw transcription — nothing else.

You'll often get a first draft that barely needs editing — because it's built from how you actually talk about the product, not how AI imagines you might. That said: always review the output. AI will drop nuances that matter, and sometimes adds things you never said.

Step 3: A Second Round of References

Yes — more references. A significant amount of production work happens before a camera rolls or animation starts.

For complex or abstract scenes — especially anything involving AI generation or motion graphics — you need more specific visual examples than the first round provided.

For the AMPD project, after setting the overall 3D direction in round one, we pulled specific Pinterest images for the AI-generated character scenes: what the characters looked like, what they were wearing, what the office environment felt like. A modern open-plan Google-style office reads completely differently from a formal corporate environment — and that difference shows up in the final video.

Here's what typically gets approved at the concept stage, depending on your video type:

Live-action shoot — location with specific decoration references; character style if actors are involved
AI visualization — same approvals as live-action: location, set decoration, characters, wardrobe
2D animation — 2–3 key frames and character designs based on approved references, to lock in the illustration and motion style
3D animation — static mockups of key scenes with materials and textures approved (colors, metals, plastics — especially critical for product visualizations)
Stock + graphic templates — for each scene in the script, find and approve the specific clip before editing begins

When a client sees the creative direction before production starts, there are no surprises at delivery. Especially for a first-time video buyer, your agency should be able to explain every visual decision clearly — before a frame is shot or generated.

Step 4: Production — Where AI Actually Fits

There are two ways AI plays a role in production: generating video footage, and supporting the process as a tool throughout.

Should you generate footage with AI?

For talking-head segments with your team — shoot those for real. Your team is a credibility asset and AI avatars, even technically impressive ones, don't carry the same weight in B2B contexts.

For everything else, the decision comes down to what kind of shots you need. On the AMPD project, we couldn't find stock footage with the right combination of character types, locations, and poses. So we generated those scenes with AI. You can see how that worked in the full AMPD case study.

In another project, we made a Pixar-style animated short with animals using AI generation — a video that simply wouldn't exist otherwise, because hand-animating at Pixar quality costs tens of thousands of dollars per second.

Where NOT to use AI

On a medical device project, we needed stable, precise shots of internal device components. AI generation produced too many glitches and continuity errors. In 3D software, we have full control over every material, angle, and detail.

The principle: if technical precision matters more than creative variety, traditional 3D tools are usually the better call. That's why at the scripting and references stage, we tell every client exactly where we'll use AI, where we'll use 3D, and where we'll use a camera.

AI in post-production

AI also runs quietly through post-production in ways you might not notice. Our team uses DaVinci Resolve for editing and color grading — it's packed with AI features, from intelligent focus corrections on already-filmed footage to reconstructing a damaged audio take from a voiceover artist.

All 4 Prompts in One Place

Copy and keep these. Use them in Claude or Gemini.

Prompt 1 — Competitor research

I'm a [role] at [Company Name]. Website: [URL] Product: [brief description] Target audience: [who they are] Market/niche: [industry or category] Geography: [e.g. US, Europe, global] Task: Find 15 direct competitors. For each one provide: — Company name — Website URL — What they sell and to whom — Any notable differentiator Sort by relevance to my business.

Prompt 2 — YouTube channel finder

Here is a list of companies: [paste list] Task: Find the official YouTube channel for each company. Return a table with two columns: — Company name — YouTube channel URL If no channel is found, mark it as "Not found".

Prompt 3 — Script draft

You are a scriptwriter specializing in B2B video production. Write a video script based on the following: Product: [name and one-line description] Key benefits: [list 2–4 main benefits] Value delivered: [what problem it solves and for whom] Target audience: [job title, company size, industry] Core message: [the one thing viewers must take away] CTA: [what you want viewers to do after watching] Video type: [explainer / product demo / testimonial / ad] Runtime: [e.g. 30 sec / 60 sec / 2 min] Tone: [e.g. professional, conversational, bold] Format the output as: — Scene-by-scene breakdown — Voiceover / on-screen text — Visual direction per scene

Prompt 4 — Raw transcription (run before Prompt 3)

Step 1 — Transcribe only. Below is my voice recording / raw notes: [paste text or attach audio] Transcribe it word for word. Do not clean up, summarize, rephrase, or remove anything. Preserve every word exactly as spoken, including filler words and repetitions. Output only the raw transcription — nothing else.

The Bottom Line

Whether you're working with a freelancer or an agency, with AI or with 3D — making a great SaaS video always starts the same way: define your audience, define your goal, and collect examples of what you want it to look and feel like. The AI tools change. This process doesn't.

If you want to see how this workflow plays out in real projects, these two case studies show the full process:

If you're ready to start your own SaaS video — get in touch with the Lava Media team. First conversation is free.